SGLang on the NVIDIA DGX Spark
On the NVIDIA DGX Spark, SGLang runs Qwen3.8 125B · 6B active at 4-bit at a median 35.0 tokens per second for generation, from 1 plain run (256K context).
Speed by model on the NVIDIA DGX Spark
Plain runs only: one device, one request, no speculative decoding, the whole model in memory. Each row is one model size at one bit class, with the other engines' plain runs of the same model on this card beside it. Context length and build still differ between runs, and each side shows its own.
| Model | SGLang | Other engines |
|---|---|---|
| Qwen3.8 125B · 6B active 4-bit | 35.0 t/sone report, 256K context | llama.cpp 24.5 t/sone report, 128K context |
5 SGLang reports on the NVIDIA DGX Spark
Newest first. A run on several cards, with speculative decoding, with part of the model in system RAM or with batched requests says so, since each describes a different setup.
- DeepSeek V4.1 Flash 552B · 16B active · MXFP487.7 t/s(4 cards, speculative decoding)
- Qwen3.8 27B · NVFP441.0 t/s(speculative decoding)
- Qwen3.8 27B · NVFP434.0 t/s(speculative decoding)
- Qwen3.8 125B · 6B active · NVFP435.0 t/s
- Qwen3.8 125B · 6B active · NVFP451.1 t/s(4 cards, speculative decoding)
Frequently asked
How fast is SGLang on the NVIDIA DGX Spark?
On the NVIDIA DGX Spark, SGLang runs Qwen3.8 125B · 6B active at 4-bit at a median 35.0 tokens per second for generation, from 1 plain run (256K context).
Is SGLang faster than other engines on the NVIDIA DGX Spark?
Same card, same model size and bit class, plain runs only: Qwen3.8 125B · 6B active at 4-bit gives SGLang 35.0 t/s (1 run) and llama.cpp 24.5 t/s (1). Context length and engine build differ between the runs, so treat it as a lead, not a benchmark.